Refractometer

library(tidyverse)

There were 13 warnings (use warnings() to see them)

refrac <- read_tsv("/Users/jkim/Downloads/refract\ data\ -\ Sheet1 (3).tsv")

Rows: 147 Columns: 7── Column specification ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Delimiter: "\t"
chr (4): Refract, Owner, Unit, Solution
dbl (3): Reading, Trial, Expected
ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

refrac %>%
  ggplot(aes(x=Refract, y=Reading, color=Owner)) +
  geom_hline(aes(yintercept = Expected)) +
  geom_point() +
  facet_wrap(~Solution, scales="free_y")

refrac %>%
  ggplot(aes(x=Refract, y=Reading, color=Owner)) +
  geom_boxplot() +
  geom_hline(aes(yintercept = Expected)) +
  facet_wrap(~Solution, scales="free_y") +
  ggtitle("Comparison of refractometers")

  
refrac %>%
  ggplot() +
  geom_boxplot(aes(x=Refract, y=Reading, fill=Refract)) +
  geom_point(aes(x=Refract, y=Reading, color=Owner)) +
  geom_point(aes(x=Refract, y=Reading), pch=21, color="white") +
  geom_hline(aes(yintercept = Expected)) +
  facet_wrap(~Solution, scales="free_y") +
  ggtitle("Comparison of refractometers")

refrac %>%
  mutate(norm_reading = Reading / Expected) %>%
  ggplot(aes(x=Refract, y=norm_reading, color=Owner)) +
  geom_hline(yintercept = 1) +
  geom_point() +
  facet_wrap(~Solution, scales="free_y")

# normalize readings
refrac %>%
  filter(!is.na(Expected)) %>%
  mutate(norm_reading = Reading / Expected) %>%
  ggplot() +
  geom_boxplot(aes(x=Refract, y=norm_reading, color=Owner))

refrac %>%
  filter(!is.na(Expected)) %>%
  filter(Solution != "2% Folgers") %>%
  mutate(norm_reading = Reading / Expected) %>%
  ggplot() +
  geom_violin(aes(x=Refract, y=norm_reading, color=Owner))

Plot the error of each refractometer at different ratios

# Center and scale
refrac %>%
  group_by(Refract, Owner, Solution) %>%
  mutate(norm_reading = (Reading - mean(Reading))/mean(Reading) * 100) %>%
  ggplot() +
  geom_density(aes(norm_reading, fill = Refract), alpha=0.25) +
  ggtitle("Center and scaled readings by refractometer")+
  xlab("Percent error")

refrac %>%
  group_by(Refract, Owner, Solution) %>%
  mutate(norm_reading = (Reading - mean(Reading))/mean(Reading)) %>%
  ggplot(aes(x=Refract, y=norm_reading, color=Owner)) +
  geom_boxplot() +
  facet_wrap(~Solution) +
  ggtitle("Centered and scaled") +
  ylab("Centered and scaled")

refrac %>%
  filter(!Solution %in% c("Filter Brew", "2% Folgers")) %>%
  group_by(Refract, Owner, Solution) %>%
  mutate(norm_reading = (Reading - mean(Reading))/mean(Reading)) %>%
  ggplot(aes(x=Refract, y=norm_reading, color=Owner)) +
  geom_boxplot() +
  facet_wrap(~Solution) +
  ggtitle("Centered and scaled") +
  ylab("Centered and scaled")

# Known values only
refrac %>%
  filter(!is.na(Expected)) %>%
  group_by(Refract, Owner, Solution) %>%
  mutate(norm_reading = (Reading - Expected)/Expected * 100) %>%
  ggplot() +
  geom_density(aes(norm_reading, fill = Refract), alpha=0.25) +
  ggtitle("Error relative to expected value")+
  xlab("Percent error")

refrac %>%
  filter(!is.na(Expected)) %>%
  group_by(Refract, Owner, Solution) %>%
  mutate(norm_reading = (Reading - Expected)/Expected * 100) %>%
  ggplot(aes(x=Refract, y=norm_reading, color=Owner)) +
  geom_boxplot() +
  facet_wrap(~Solution) +
  ggtitle("Comparison of refractometers") +
  ylab("Percent error")

refrac %>%
  filter(!is.na(Expected)) %>%
  filter(Solution != "2% Folgers") %>%
  group_by(Refract, Owner, Solution) %>%
  mutate(norm_reading = (Reading - Expected)/Expected * 100) %>%
  ggplot(aes(x=Refract, y=norm_reading, color=Owner)) +
  geom_boxplot() +
  facet_wrap(~Solution) +
  ggtitle("Comparison of refractometers") +
  ylab("Percent error")

refrac %>%
  filter(!is.na(Expected)) %>%
  filter(Solution != "2% Folgers") %>%
  group_by(Refract, Owner, Solution) %>%
  mutate(norm_reading = (Reading - Expected)/Expected * 100) %>%
  ggplot() +
  geom_density(aes(norm_reading, fill = Refract), alpha=0.25) +
  ggtitle("No 2% Folgers")+
  xlab("Percent error")

# Let's see how the errors propogate to EY

ey_key <- tribble(
  ~Solution, ~ratio, ~type,
  "9.95% Folgers", 2, "9.95% Folgers",
  "9.95% Folgers 2", 2, "9.95% Folgers",
  "2% Folgers", 10, "2% Folgers"
  
)

refrac %>%
  filter(Solution %in% c("9.95% Folgers", "9.95% Folgers 2", "2% Folgers")) %>%
  left_join(ey_key, by="Solution") %>%
  mutate(predicted_ey = Reading * ratio) %>%
  mutate(expected_ey = Expected * ratio) %>%
  mutate(diff_from_expected = predicted_ey - expected_ey) %>%
  ggplot() +
  geom_density(aes(diff_from_expected, fill=Refract), alpha=0.25) +
  ggtitle("Error in Extraction Yield (%), Folgers Instant") +
  xlab("Observed EY - Expected EY")

refrac %>%
  filter(Solution %in% c("9.95% Folgers", "9.95% Folgers 2", "2% Folgers")) %>%
  left_join(ey_key, by="Solution") %>%
  mutate(predicted_ey = Reading * ratio) %>%
  mutate(expected_ey = Expected * ratio) %>%
  mutate(diff_from_expected = predicted_ey - expected_ey) %>%
  ggplot() +
  geom_point(aes(x=expected_ey, y=predicted_ey, color=Refract)) +
  ggtitle("Error in Extraction Yield(%), Folgers Instant")

ratio_expansion <- ey_key %>%
  expand(Solution, ratio=5:15) %>%
  left_join(select(ey_key, Solution, type))

Joining, by = "Solution"

refrac %>%
  filter(Solution %in% c("2% Folgers")) %>%
  left_join(ratio_expansion, by="Solution") %>%
  mutate(predicted_ey = Reading * ratio) %>%
  mutate(expected_ey = Expected * ratio) %>%
  mutate(diff_from_expected = predicted_ey - expected_ey) %>%
  group_by(expected_ey, Refract, ratio) %>%
  summarise(mean_obs = mean(predicted_ey),
            sd_high_obs = mean(predicted_ey) + sd(predicted_ey * 2),
            sd_low_obs = mean(predicted_ey) - sd(predicted_ey * 2)) %>%
  ungroup() %>%
  ggplot(aes(x=expected_ey, y=mean_obs, color=Refract)) +
  geom_line() +
  geom_ribbon(aes(ymin=sd_low_obs, ymax=sd_high_obs, fill=Refract), alpha=0.25, color=NA) +
  geom_abline() +
  ggtitle("Error in Extraction Yield(%), Folgers Instant 2% TD, 2 sd ribbon")

`summarise()` has grouped output by 'expected_ey', 'Refract'. You can override using the `.groups` argument.

refrac %>%
  filter(Solution %in% c("2% Folgers")) %>%
  left_join(ratio_expansion, by="Solution") %>%
  mutate(predicted_ey = Reading * ratio) %>%
  mutate(expected_ey = Expected * ratio) %>%
  mutate(diff_from_expected = predicted_ey - expected_ey) %>%
  group_by(Owner, expected_ey, Refract, ratio) %>%
  summarise(mean_obs = mean(predicted_ey),
            sd_high_obs = mean(predicted_ey) + sd(predicted_ey * 2),
            sd_low_obs = mean(predicted_ey) - sd(predicted_ey * 2)) %>%
  ungroup() %>%
  mutate(instrument = paste(Refract, Owner)) %>%
  ggplot(aes(x=expected_ey, y=mean_obs, color=instrument)) +
  geom_ribbon(aes(ymin=sd_low_obs, ymax=sd_high_obs, fill=instrument), alpha=0.25, color=NA) +
  geom_line() +
  geom_abline() +
  ggtitle("Error in Extraction Yield(%), Folgers Instant 2% TDS, 2 sd ribbon") +
  facet_wrap(~Refract, ncol=1)

`summarise()` has grouped output by 'Owner', 'expected_ey', 'Refract'. You can override using the `.groups` argument.

refrac %>%
  filter(Solution %in% c("2% Folgers")) %>%
  left_join(ratio_expansion, by="Solution") %>%
  mutate(predicted_ey = Reading * ratio) %>%
  mutate(expected_ey = Expected * ratio) %>%
  mutate(diff_from_expected = predicted_ey - expected_ey) %>%
  group_by(Owner, expected_ey, Refract, ratio) %>%
  summarise(med_obs = median(predicted_ey),
            high_obs = max(predicted_ey),
            low_obs = min(predicted_ey)) %>%
  ungroup() %>%
  mutate(instrument = paste(Refract, Owner)) %>%
  ggplot(aes(x=expected_ey, y=med_obs, color=instrument)) +
  geom_ribbon(aes(ymin=low_obs, ymax=high_obs, fill=instrument), alpha=0.25, color=NA) +
  geom_line() +
  geom_abline() +
  ggtitle("Error in Extraction Yield(%), Folgers Instant 2% TDS, min/max ribbon") +
  facet_wrap(~Refract)

`summarise()` has grouped output by 'Owner', 'expected_ey', 'Refract'. You can override using the `.groups` argument.

ratio_expansion <- ey_key %>%
  expand(Solution, ratio=(15:28)/10) %>%
  left_join(select(ey_key, Solution, type))

Joining, by = "Solution"

refrac %>%
  filter(Solution %in% c("9.95% Folgers", "9.95% Folgers 2")) %>%
    left_join(ratio_expansion, by="Solution") %>%
  mutate(predicted_ey = Reading * ratio) %>%
  mutate(expected_ey = Expected * ratio) %>%
  mutate(diff_from_expected = predicted_ey - expected_ey) %>%
  group_by(expected_ey, Refract, ratio) %>%
  summarise(mean_obs = mean(predicted_ey),
            sd_high_obs = mean(predicted_ey) + sd(predicted_ey) * 2,
            sd_low_obs = mean(predicted_ey) - sd(predicted_ey) * 2) %>%
  ungroup() %>%
  ggplot(aes(x=expected_ey, y=mean_obs, color=Refract)) +
  geom_line() +
  geom_abline() +
  geom_ribbon(aes(ymin=sd_low_obs, ymax=sd_high_obs, fill=Refract), alpha=0.25, color=NA) +
  ggtitle("Error in Extraction Yield(%), Folgers Instant 9.95% TDS, 2 sd ribbon")

`summarise()` has grouped output by 'expected_ey', 'Refract'. You can override using the `.groups` argument.

refrac %>%
  filter(Solution %in% c("9.95% Folgers", "9.95% Folgers 2")) %>%
    left_join(ratio_expansion, by="Solution") %>%
  mutate(predicted_ey = Reading * ratio) %>%
  mutate(expected_ey = Expected * ratio) %>%
  mutate(diff_from_expected = predicted_ey - expected_ey) %>%
  group_by(Owner, expected_ey, Refract, ratio) %>%
  summarise(mean_obs = mean(predicted_ey),
            sd_high_obs = mean(predicted_ey) + sd(predicted_ey) * 2,
            sd_low_obs = mean(predicted_ey) - sd(predicted_ey) * 2) %>%
    mutate(instrument = paste(Refract, Owner)) %>%
  ggplot(aes(x=expected_ey, y=mean_obs, color=instrument)) +
  geom_ribbon(aes(ymin=sd_low_obs, ymax=sd_high_obs, fill=instrument), alpha=0.25, color=NA) +
  geom_line() +
  geom_abline() +
  ggtitle("Error in Extraction Yield(%), Folgers Instant 9.95% TDS, 2 sd ribbon") +
  facet_wrap(~Refract, ncol=1)

`summarise()` has grouped output by 'Owner', 'expected_ey', 'Refract'. You can override using the `.groups` argument.

refrac %>%
  filter(Solution %in% c("9.95% Folgers", "9.95% Folgers 2")) %>%
  left_join(ratio_expansion, by="Solution") %>%
  mutate(predicted_ey = Reading * ratio) %>%
  mutate(expected_ey = Expected * ratio) %>%
  mutate(diff_from_expected = predicted_ey - expected_ey) %>%
  group_by(Owner, expected_ey, Refract, ratio) %>%
  summarise(med_obs = median(predicted_ey),
            high_obs = max(predicted_ey),
            low_obs = min(predicted_ey)) %>%
  ungroup() %>%
  mutate(instrument = paste(Refract, Owner)) %>%
  ggplot(aes(x=expected_ey, y=med_obs, color=instrument)) +
  geom_ribbon(aes(ymin=low_obs, ymax=high_obs, fill=instrument), alpha=0.25, color=NA) +
  geom_line() +
  geom_abline() +
  ggtitle("Error in Extraction Yield(%), Folgers Instant 9.95% TDS, min/max ribbon") +
  facet_wrap(~Refract, ncol=1)

`summarise()` has grouped output by 'Owner', 'expected_ey', 'Refract'. You can override using the `.groups` argument.